用图神经网络加速高能物理中的图分类计算，提升效率并可解释。
Graph Neural Networks for the Graphical Bootstrap
- 基于图神经网络处理超2000万张图的分类任务
- 在99.996%的ROC AUC下实现对大图的鲁棒泛化
- 压缩冗余数据达85.5%，适合高能物理与复杂系统研究者
我们研究了一个涉及超过2000万张图的图分类问题，源于平面$/mathcal{N}=4$超杨-米尔斯理论中关联子的高阶微扰计算，该理论与强核力理论密切相关。我们基准测试了图神经网络，包括图变压器，在高达99.996%的ROC AUC下实现了对更大图的稳健泛化。随后，分析了模型如何通过将分母图的冗余数据减少高达85.5%，相较传统图形自举算法实现计算加速。最后，我们研究了模型嵌入以探究其可解释性。
原文摘要 · Abstract (English)
We study a graph classification problem involving over 20 million graphs, arising from high-order perturbative computations of correlators in planar $\mathcal{N}=4$ super-Yang--Mills, a model closely related to the theory of the strong nuclear force. We benchmark graph neural networks, including graph transformers, achieving robust generalization to larger graphs with up to $99.996\%$ ROC AUC. Then, we analyze how the models can be used to gain a computational speedup compared to the traditional graphical bootstrap algorithm, through shrinking the redundant data by up to $85.5\%$ at the level of denominator graphs. Finally, we study the embeddings of the models to investigate their interpretability.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。